mlsanityData to 7 Oct 2026

Self-Training for End-to-End Speech Recognition

arXiv 1909.09116 ↗First posted 19 Sep 2019cs.CL cs.LG eess.AScited by 128

Not on the map. Why?

Abstract (arXiv)

We revisit self-training in the context of end-to-end speech recognition. We demonstrate that training with pseudo-labels can substantially improve the accuracy of a baseline model. Key to our approach are a strong baseline acoustic and language model used to generate the pseudo-labels, filtering mechanisms tailored to common errors from sequence-to-sequence models, and a novel ensemble approach to increase pseudo-label diversity. Experiments on the LibriSpeech corpus show that with an ensemble of four models and label filtering, self-training yields a 33.9% relative improvement in WER compared with a baseline trained on 100 hours of labelled data in the noisy speech setting. In the clean speech setting, self-training recovers 59.3% of the gap between the baseline and an oracle model, which is at least 93.8% relatively higher than what previous approaches can achieve.

Key facts. Cited by 128 papers in the map (data through 7 Oct 2026). First posted on arXiv 19 Sep 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 40; 2022: 31; 2023: 26; 2024: 17; 2025: 3; 2026: 11.

128 papers.

Sort:

By area:

  1. 2021HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Unitscited by 2,301
  2. 2021VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretationcited by 455
  3. 2021W2v-BERT: Combining Contrastive Learning and Masked Language Modeling for Self-Supervised Speech Pre-Trainingcited by 340
  4. 2023Google USM: Scaling Automatic Speech Recognition Beyond 100 Languagescited by 228
  5. 2021Unsupervised Speech Recognitioncited by 147
  6. 2022Unsupervised Prompt Learning for Vision-Language Modelscited by 125
  7. 2023Auto-AVSR: Audio-Visual Speech Recognition with Automatic Labelscited by 113
  8. 2021BigSSL: Exploring the Frontier of Large-Scale Semi-Supervised Learning for Automatic Speech Recognitioncited by 88
  9. 2022Self-Training: A Surveycited by 85
  10. 2022Unified Speech-Text Pre-training for Speech Translation and Recognitioncited by 64
  11. 2021Scalable and Adaptive Graph Neural Networks with Self-Label-Enhanced trainingcited by 54
  12. 2022Enhanced Direct Speech-to-Speech Translation Using Self-supervised Pre-training and Data Augmentationcited by 43
38 more
  1. 2021An Exploration of Self-Supervised Pretrained Representations for End-to-End Speech Recognitioncited by 39
  2. 2021Momentum Pseudo-Labeling for Semi-Supervised Speech Recognitioncited by 37
  3. 2021Large-Scale Self- and Semi-Supervised Learning for Speech Translationcited by 36
  4. 2021Pseudo-labeling for Scalable 3D Object Detectioncited by 32
  5. 2021FlipDA: Effective and Robust Data Augmentation for Few-Shot Learningcited by 32
  6. 2023Making More of Little Data: Improving Low-Resource Automatic Speech Recognition Using Data Augmentationcited by 30
  7. 2021Pseudo-Labeling for Massively Multilingual Speech Recognitioncited by 23
  8. 2021Kaizen: Continuously improving teacher using Exponential Moving Average for semi-supervised speech recognitioncited by 21
  9. 2022A Survey of Multilingual Models for Automatic Speech Recognitioncited by 21
  10. 2021Margin Preserving Self-paced Contrastive Learning Towards Domain Adaptation for Medical Image Segmentationcited by 20
  11. 2022COSMOS: Cross-Modality Unsupervised Domain Adaptation for 3D Medical Image Segmentation based on Target-aware Domain Translation and Iterative Self-Trainingcited by 20
  12. 2024Advancing Weakly-Supervised Audio-Visual Video Parsing via Segment-wise Pseudo Labelingcited by 20
  13. 2022SPIRAL: Self-supervised Perturbation-Invariant Representation Learning for Speech Pre-Trainingcited by 19
  14. 2021Self-training Improves Pre-training for Few-shot Learning in Task-oriented Dialog Systemscited by 18
  15. 2022How does unlabeled data improve generalization in self-training? A one-hidden-layer theoretical analysiscited by 18
  16. 2023Improving Audio-Visual Video Parsing with Pseudo Visual Labelscited by 18
  17. 2021Wav-BERT: Cooperative Acoustic and Linguistic Representation Learning for Low-Resource Speech Recognitioncited by 16
  18. 2022Masked Unsupervised Self-training for Label-free Image Classificationcited by 14
  19. 2022Supervision-Guided Codebooks for Masked Prediction in Speech Pre-trainingcited by 14
  20. 2024Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputscited by 14
  21. 2021EAT: Enhanced ASR-TTS for Self-supervised Speech Recognitioncited by 13
  22. 2021Robust Mutual Learning for Semi-supervised Semantic Segmentationcited by 13
  23. 2021MDD-Eval: Self-Training on Augmented Data for Multi-Domain Dialogue Evaluationcited by 13
  24. 2021Task-adaptive Pre-training and Self-training are Complementary for Natural Language Understandingcited by 12
  25. 2021Magic dust for cross-lingual adaptation of monolingual wav2vec-2.0cited by 12
  26. 2021Contrastive Semi-supervised Learning for ASRcited by 11
  27. 2021XLST: Cross-lingual Self-training to Learn Multilingual Representation for Low Resource Speech Recognitioncited by 11
  28. 2022Continuous Pseudo-Labeling from the Startcited by 11
  29. 2024Enhancing Low-Resource ASR through Versatile TTS: Bridging the Data Gapcited by 11
  30. 2021The History of Speech Recognition to the Year 2030cited by 10
  31. 2022Improving Mispronunciation Detection with Wav2vec2-based Momentum Pseudo-Labeling for Accentedness and Intelligibility Assessmentcited by 10
  32. 2022Boosting Cross-Domain Speech Recognition with Self-Supervisioncited by 10
  33. 2021On-the-Fly Aligned Data Augmentation for Sequence-to-Sequence ASRcited by 9
  34. 2023Master-ASR: Achieving Multilingual Scalability and Low-Resource Adaptation in ASR with Modular Learningcited by 9
  35. 2024Useful Blunders: Can Automated Speech Recognition Errors Improve Downstream Dementia Classification?cited by 9
  36. 2021Leveraging pre-trained representations to improve access to untranscribed speech from endangered languagescited by 8
  37. 2021Comparing the Benefit of Synthetic Training Data for Various Automatic Speech Recognition Architecturescited by 8
  38. 2021Back-Training excels Self-Training at Unsupervised Domain Adaptation of Question Generation and Passage Retrievalcited by 8

50 most cited of 128 All 128 (JSON)